基于子图特征组合的人脸识别技术研究

Research on Face Recognition Technology Based on Sub-Image Feature Combination

提出了一种基于子图特征组合的人脸特征提取方法,并结合BP神经网络给出一种人脸识别模型.模型首先将人脸图片分割为子图,然后对每个子图进行离散余弦变换并选择最大的余弦系数表示该子图,最后将这些系数组合为向量作为整幅图像的特征.我们选择BP神经网络作为人脸识别模型中的分类器,并通过实验优化相关参数.基于ORL数据库的模拟实验表明,所提出的特征提取算法是有效的,并且模型具有较高的识别率.

Based on sub-image feature combination, a feature extraction method of face image was proposed. Combined with BP neural network, a face recognition model was presented. In the model, face images were divided into sub-images firstly, then the discrete cosine transform was operated on each sub-image and the largest coefficient was selected to denote the sub-image, finally these coefficients were combined as the feature of the whole face image. BP neural network was used as classifier in the model and its parameters were optimized according to experiments. Based on ORL face database, experimental results show that the proposed feature extraction method is effective and the model has high accuracy.